Three Index Checks That Tell You If a Paid Agent Service Has Real Customers
Multi-payer count, corroboration, and on-chain transfers answer different questions about whether an agent service has real customers. Here is how to run each check, what each one proves, and what to do when they conflict.
· 716 words
Deciding whether to integrate a paid agent service is easier when you know how many other buyers have made the same decision. The index exposes three independent signals for this, each measuring something different. Running all three takes five minutes; skipping any of them leaves a question open that can surface later as a production dependency on a service that turns out to be a single-customer operation.
Check one: the multi-payer count
The commerce dataset tracks distinct paying buyers per endpoint over the trailing 30 days. Right now, 12,434 paid endpoints have recorded more than one distinct paying buyer. This is Coinbase's call telemetry, which means the caveat travels with it: the data reflects Coinbase's network and does not cover payment flows outside that network.
What it proves: the service has at least one customer who is not the operator. A second buyer is the minimum evidence that an independent party evaluated the service and found it worth paying for. It does not prove the service is good; it proves someone other than the builder paid for it.
What it does not prove: the buyer count at the low end is a filter, not a ranking. Once a service clears the threshold, look at what those buyers actually purchased.
How to run it: query bazaar_multi_payer in the index for the specific endpoint host, or use the index API to filter endpoints by payer_count > 1.
Check two: corroboration
A separate signal: is the service listed in more than one independent registry? Currently, 1,363 hosts appear in two or more independent registries. Self-listings are excluded from this count, so what remains is services that a different operator's registry picked up independently.
What it proves: the service is real enough to have been noticed by more than one discovery system. A service that only appears in the registry where its operator published it has not yet been independently verified to exist.
What it does not prove: corroboration says nothing about whether the service works, whether it charges a fair price, or whether its customers are satisfied. It is an existence check, not a quality check.
Combined with the multi-payer check: a corroborated service with multiple buyers has cleared two independent thresholds. Neither required the operator's cooperation to achieve. A service that is not corroborated but has multiple buyers may simply be undiscovered by secondary registries; the buyer count is the stronger signal in that case.
Check three: on-chain transfers
The most verifiable signal. In the trailing week, 321,838 USDC transfers reached agent payment addresses across 1,362 distinct services. We collect this by scanning Base mainnet USDC Transfer logs directly; no external publisher is the source.
What it proves: a wallet spent money and another wallet received it. This is the only check of the three that cannot be faked by submitting data to a registry or calling your own endpoint. On-chain transfers are a settled ledger entry.
What it does not prove: a service can appear in on-chain transfer data because it is being paid by a single buyer in high volume. Concentration tells you about market structure, not about whether there are multiple independent customers. Look at the number of sending addresses, not just the transfer count.
The threshold gap to know: 1,362 services have on-chain transfers this week against 12,434 with multi-payer status in the commerce data. The two datasets measure different things at different frequencies and through different channels. A service that has multi-payer status but no on-chain transfers may use a non-USDC payment path or may be on a chain the scanner does not currently cover. A service with on-chain transfers but no multi-payer status in the commerce data should prompt a question about what is funding those transfers.
When the three checks conflict
If a service has on-chain transfers but is not corroborated and has no multi-payer status, treat the on-chain data as your floor: money is moving, but you do not yet know from whom or to what. If a service has multi-payer status and corroboration but no on-chain transfers, it may have real customers paying through a channel outside the scan. In both cases, ask for a payment address. An address converts "I have customers" from a self-declaration into something you can verify without the service's cooperation.